determined  by determined-ai

Open-source ML platform for simplifying training, tuning, tracking, and resource management

Created 5 years ago
3,187 stars

Top 15.1% on SourcePulse

GitHubView on GitHub
Project Summary

Determined AI is an open-source platform designed to streamline deep learning workflows for researchers and engineers. It simplifies distributed training, hyperparameter tuning, experiment tracking, and resource management, aiming to accelerate model development and reduce cloud costs.

How It Works

Determined operates through a Python library, a command-line interface (CLI), and a web UI. The Python library allows users to integrate their PyTorch or TensorFlow code by organizing it within PyTorchTrial or TFv1Trial classes or by using the lower-level Core API. This abstraction handles distributed execution and hyperparameter search, enabling faster training and automated tuning. The CLI manages cluster deployment (local, AWS, GCP) and experiment execution, while the Web UI provides comprehensive visualization of training progress, resource utilization, and model artifacts.

Quick Start & Requirements

  • Install CLI: pip install determined
  • Deployment: Use det deploy local cluster-up for local setup or det deploy aws up / det deploy gcp up for cloud.
  • Prerequisites: Python, Docker (for local deployment). Cloud deployment requires cloud provider credentials.
  • Resources: Cloud GPU instances are recommended for training.
  • Links: Examples, Documentation, Quick Start Guide

Highlighted Details

  • Supports distributed training across multiple GPUs and nodes.
  • Integrates advanced hyperparameter search algorithms like Adaptive SHA.
  • Provides robust experiment tracking, including code snapshots and configuration history.
  • Offers performance profiling and resource management for cost optimization.

Maintenance & Community

Health Check
Last Commit

6 months ago

Responsiveness

1 day

Pull Requests (30d)
0
Issues (30d)
0
Star History
11 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.